Cluster Analysis Based on Pre-specified Multiple Layer Structure
Cluster analysis can be divided into two categories; hierarchical and non-hierarchical cluster analyses. In the present study, a method of cluster analysis which does not utilize hierarchical nor non-hierarchical procedures is introduced. The present cluster analysis pre-specifies a structure having multiple layers, e.g., the species, the genus, the family, and the order. The highest layer or layer 0 has one cluster which all objects belong to. The cluster at layer 0 has the pre-specified number of clusters at the next lower layer or layer 1. Each cluster at layer 1 has the pre-specified number of clusters at the next lower layer or layer 2, and so on. The cluster analysis classifies the object into one of the clusters at all layers simultaneously. While the cluster structure is hierarchical, the procedure is not hierarchical which is different from that of the agglomerative or divisive algorithms of the hierarchical cluster analysis. The algorithm tries to optimize the fitness measure at all layers simultaneously. The cluster analysis is applied to the data on whisky molts.
- Arabie, P., & Hubert, L. (1994). Cluster analysis in marketing research. In R. P. Bgozzi (Ed.), Advanced methods of marketing research (pp. 160–189). Cambridge: Blackwell.Google Scholar
- Coombs, C. H. (1964). A theory of data. New York: Wiley.Google Scholar
- MacQueen, J. B. (1967). Some methods for classification and analysis of multivariate observations. In Proceeding of the 5th Berkeley symposium on mathematical statistics and probability, Berkeley (Vol. 2, PP. 281–297).Google Scholar
- Okada, A., & Yokoyama, S. (2010). Multi-layer cluster analysis. In Proceeding of the 28th meeting of the Japanese classification society, Tokyo, Japan (pp. 11–12).Google Scholar
- Wishart, D. (2002). Whisky classified: Choosing single molts by flavours. London: Pavillion.Google Scholar